Agents
Agents are the core execution unit of Procurator — persistent, governable AI entities that combine a language model, a behavioral system prompt, tool access via MCP, and a curated knowledge base into a single deployable unit.
Agents Registry — manage, configure, and monitor your autonomous agent fleet from a single table.
How It Works
When an agent is invoked, Procurator orchestrates the following pipeline:
- Register — Create an agent from a profile template that pre-populates model, system instructions, and governance defaults.
- Arm — Attach MCP tool servers, assign knowledge bases for RAG retrieval, and configure generation parameters on dedicated config tabs.
- Govern — Set PII redaction, security and compliance tiers, IAM role allowlists, token budgets, and pre/post execution hooks.
- Deploy — Invoke via Chat playground, REST API, Workflow steps, Teams, or scheduled cron runs. Every execution is logged and attributed to FinOps.
Key Capabilities
Agent Registry
A central catalog of all agents in your organization, filterable by model, status, team membership, and creation date.
Tool Mounting
Attach MCP servers to grant agents access to web search, code execution, database queries, RAG ingestion, and custom APIs.
Streaming Execution
Run agents with real-time output streaming via SSE. Watch thoughts, tool calls, and responses appear live in the Chat interface.
PII Redaction & Security
Per-agent PII enforcement, classification tiers, pre- and post-execution hooks, and production approval gating.
Governance & FinOps
Generation tuning sliders, concurrent execution caps, per-agent token budgets, and optimization routing policies.
Multi-Agent Supervision
Designate any agent as a Team Supervisor, enabling it to coordinate sub-agents in orchestrated multi-agent workflows.
Ready to build your first agent?
Full configuration reference, profile templates, field definitions, and API docs are in the developer documentation.
Read the Docs →